Critical periods and adult language learning: what closes and what stays open

The Short Version

Some windows really do narrow. Within a couple of years of birth, babies lose sensitivity to speech sounds their language doesn't use, and speech has a stronger critical period than most skills, which is why a child picks up a second language, and its accent, more easily than an adult starting from scratch [1][2]. But the sources are just as clear that the adult brain stays changeable: humans keep speech circuits in a relatively juvenile state for life, adults who already speak several languages learn new ones more easily, and Andrew Huberman argues that the old idea of a permanently shut critical period was "quite wrong" [2][3]. What changes after childhood, and especially after about age 25, is the price of admission: adults can't learn by passive exposure and instead need alertness, focused attention, small increments, many errors, a real reason to learn, and sleep afterward [3][4]. The realistic message for adult learners is not "too late," but "different rules," with native-like sound systems the hardest part to reach.

Table of contents

What They Said

Eddie Chang: the ear tunes itself early, and the environment helps set the clock

Eddie Chang, chair of neurosurgery at UCSF, whose lab studies how the brain hears and produces speech, began his career studying critical periods in the auditory cortex as a medical student with neuroplasticity pioneer Mike Merzenich [1]. Early in life, he explains, there is a critical or sensitive period when the brain is highly shaped by what it hears, and the human version is striking: "after a couple of years you lose sensitivity to sounds that are not part of your native language and you have high sensitivity for the languages of your native culture" [1]. He calls it "pretty extraordinary that human brain has that flexibility yet at the same time has that specialization for language" [1].

His rat experiments show that the timing of this window is not fixed by genes alone. When he raised rat pups in continuous, moderate white noise that masked environmental sounds, the critical period, the time when the auditory cortex is "open to plasticity, it's open to change, it's open to reorganization," stayed open "much, much longer," for months beyond when it normally closes [1]. That might sound like a gift, but Chang says "it's also a retardation. It's actually slowed the maturation of auditory cortex" [1]. His conclusion: "it's not just about the genetic programming that specifies some of this sensitive period, but it's also a little bit about the nature of the sounds that we hear" that keeps the window open or closes it [1]. In other words, structured sound is what the young brain uses to mature, and the closing of the window is part of becoming specialized [1]. Asked about white noise machines for infants, he says the studies haven't been done, but he did not use white noise with his own three children and suggests more natural soothing sounds [1].

For language learning specifically, Chang offers three principles: earlier and more immersive is better; exposure must be maintained or it can be lost; and real human interaction matters more than recordings [1]. His own life adds a note of reassurance about bilingual brains: Chang speaks English and Chinese, and the same brain areas largely process both languages, though what they do with the sequences of sounds can differ [1].

Chang's work also shows why accents are so sticky. Speaking is a precise motor skill, "shaping the breath," with the larynx, pharynx, mouth, tongue and lips turning vibration into consonants and vowels [1]. About 12 articulatory movements combine to create every word, and English uses about 40 phonemes while Hawaiian uses about 12 to 14 [1]. The precentral gyrus, part of the motor cortex controlling lips, jaw and larynx, turns out to be crucial for formulating and expressing words [1]. Foreign accent syndrome makes the point in reverse: damage to that area can make someone sound as if they have a different accent without actually knowing the language [1]. An accent, on this view, lives largely in finely tuned motor and auditory maps laid down early.

Erich Jarvis: speech has a stronger critical period, but we stay "juvenile"

Erich Jarvis, who studies vocal learning in songbirds, parrots and humans at Rockefeller University, confirms the pattern from comparative biology [2]. Critical periods are one of the behavioral parallels between human speech and birdsong [2]. He points to the sad evidence from feral children: when a child isn't raised with humans until puberty, it "becomes hard for them to learn a language as an adult. So there's this critical period where you learn best. And even later on, when you're in regular society, it's hard to learn" [2]. Vocal learners also depend on hearing throughout life: when humans go deaf, "our speech starts to deteriorate without any kind of therapy," while a deaf chicken or monkey changes very little [2].

Huberman put the adult learner's question to him directly. His 14-year-old niece speaks Spanish and English fluently; "I've tried to learn Spanish as an adult. It's really challenging" [2]. He asked: "is it easier to learn multiple languages without an accent early in life?" and, if you already speak more than one language, is it easier to acquire new ones later? Jarvis answered: "So the answer to both of those questions is yes" [2].

His explanation is careful. First, the critical period is not unique to language: "the entire brain is undergoing a critical period development, not just the speech pathways," so "it's easier to learn how to play a piano. It's easier to learn how to ride a bike" as a child [2]. He specifies what "easier" means: "easier in terms of when you start from first principles of learning something," for example learning Chinese for the very first time as a child versus as an adult [2]. But "speech behavior I think has a stronger critical period" than other circuits [2].

Why close a window at all? Jarvis's answer is capacity and stability: "the brain can only hold so much information," rapid learning requires dumping old information, and "for survival you don't want to keep forgetting things," so the brain solidifies the circuits built in childhood and uses them for life [2]. Yet humans are unusual. An extra copy of a gene called SRGAP2 keeps our speech circuits and other brain regions in a more immature state throughout life than in other animals: "We're still juvenile like compared to other animals," he says, though "we still go through the critical periods like they all do" [2].

The multilingual advantage, in Jarvis's account, is not that some brains are more plastic; it is about the sound inventory. We're born able to produce a set of innate phonemes and narrow them down to the ones our language uses, and we keep those as adults [2]. When a new language uses phonemes, or combinations, you aren't used to, "it's like starting from first principles," but if you already use them across multiple languages, a third or fourth language is easier [2]. He is honest that the field isn't unanimous: this is "a common finding out there in the literature. There's some that argue against it" [2].

Jarvis also explains why children raised with two languages blend them. Describing a child whose parents speak Chinese and English, he says "cultural evolution remarkably tracks genetic evolution": the child can merge phonemes and words from both because they experience both languages at once during their critical period years, in a way adults can't [2]. His songbird work adds a social dimension: a young zebra finch prefers to learn from its own species but, if that tutor is removed, learns a Bengalese finch's song very well, pointing to social bonding as part of vocal learning [2]. Marler called the built-in bias "the innate predisposition to learn" [2].

For adults, Jarvis's practical advice is about keeping these circuits in use. He sees texting as "use it or lose it," and because speech circuits sit next to movement circuits, he argues "if you want to stay cognitively intact until your old age, you better be moving," recommending consistent dancing, walking or running plus practicing speech, oratory or singing, while noting this conviction comes from personal experience [2]. His lab also found that adults can improve a stutter with behavioral therapy, like speaking more slowly or tapping out a rhythm, which he thinks works through matching what you hear to what you produce [2], a reminder that adult speech motor patterns can still be retrained.

Andrew Huberman: the child's plasticity is free, the adult's is earned

Huberman's solo episodes explain what actually changes around adulthood [3][4]. In childhood, "we can learn through almost passive experience. We don't have to focus that hard in order to learn new things" [3]. Children go from no language to sentences that include words they've never heard, because speech and language areas are "primed to learn and create new combinations" [3]. "What this tells us is that the young brain is a plasticity machine, but then right about age 25 plus or minus a year or two, everything changes" [3]. After that, "we have to engage in a completely different set of processes" for changes to occur and to stick around [3].

Part of the reason is structural. After puberty, around 14 or 15, the human brain adds very few, if any, new neurons, and adult change comes from strengthening and removing connections [3]. Young brains have space between neurons to sample connections, while older brains fill that space with extracellular matrix and glial cells, "kind of like pouring concrete between rocks," which makes rewiring harder [3].

Huberman tells the history of the critical period idea and its correction. David Hubel and Torsten Wiesel won a Nobel Prize for showing that if one eye was closed during a critical period early in life, the visual brain rewired to represent only the open eye [3]. They believed that missing the window meant no rescue later, and Huberman, who calls them his scientific great-grandparents, says plainly: "they were quite wrong about this critical period thing" [3]. In the early 1990s, Greg Recanzone in Merzenich's lab had adults detect subtle changes in the spacing of bumps on a spinning drum; as they paid close attention, their brain's finger map changed rapidly [3]. In the key control, people touching the same drum but attending to a tone changed hearing areas instead [3]. "The experiences that you pay super careful attention to are what open up plasticity," he says [3]. And that, he argues, is "the opportunity for any of us to change our brain at any point throughout our lifespan, essentially for anything that we want to learn," including "learning new information when learning a new language" [3].

But he is equally blunt about what doesn't work. "One of the biggest lies in the universe that seems quite prominent right now is that every experience you have changes your brain" [3]. Adults change only when certain neurochemicals are present while specific neurons are active [3]. The gate has three parts: epinephrine for alertness ("So, no alertness, no neuroplasticity"), acetylcholine from the brainstem that sharpens signal over noise ("Acetylcholine acts as a spotlight"), and acetylcholine from the nucleus basalis, which in experiments made whatever the subject attended to take over part of the brain in essentially one trial [3]. With all three, he says, the nervous system "has to change" [3]. He adds that you can't simply listen to material while asleep and learn it [3].

The rewiring itself happens later: plasticity "doesn't occur during wakefulness. It occurs during sleep" [3]. Focused learning marks synapses; sleep that night and on following nights strengthens them [3]. He cites a Cell Reports paper where a 20-minute non-sleep deep rest session or shallow nap right after a hard sequence-memory task improved learning significantly [3]. His summary: "The key to plasticity in childhood is to be a child"; for adults it takes deliberate, focused effort followed by real rest [3].

Huberman on errors, increments and stakes

In a second episode, Huberman turns to how adults can recapture large changes, using Eric Knudsen's prism-glasses experiments [4]. Prisms shifted subjects' visual world so they reached for things in the wrong place. Young subjects adjusted within a day or two; "In adults, it tends to be very slow and most individuals never actually accomplish the full map shift" [4]. That is the adult critical-period problem in miniature.

Knudsen's lab found two ways around it [4]. The first is incremental learning: shifting the world by about 7 degrees, then 14, then 28. "The adult nervous system can tolerate smaller and smaller errors over time, but that you can stack those errors so that you can get a lot of plasticity" [4]. "Put simply incremental learning as an adult is absolutely essential," and "It's a mistake to try and learn a lot of information in one learning bout as an adult" [4]. The second is high contingency: when adult subjects had to use the shifted cues to find food, their plasticity became as dramatic as a young subject's [4]. "How badly we need or want the plasticity determines how fast that plasticity will arrive" [4]. He uses language as the example: "if you really need to learn conversational French to save your relationship the chances are you're going to learn it" [4]. There are limits, though; no pressure could make him learn conversational French "in the next 120 seconds" [4].

Errors are the signal. "And so errors are the basis for neuroplasticity and for learning" [4]. Frustration reflects epinephrine and acetylcholine focusing attention on the gap between what you're doing and what you want, and dopamine arrives as you start to get it right [4]. He recommends learning in roughly 90-minute bouts, with focus usually kicking in around 10 to 15 minutes, and staying in the error zone for about 7 to 30 minutes [4]. Telling yourself errors are good for learning attaches dopamine to them and speeds change [4]. He recommends arriving at the right state, too alert calls for a physiological sigh or panoramic vision, too tired calls for sleep, NSDR, caffeine or faster breathing, a problem he calls limbic friction [4].

He adds one provocative idea that bears on adult learners. Children move constantly in many relationships to gravity, and balance errors cause the cerebellum to signal deep brain centers that release dopamine, epinephrine and acetylcholine [4]. "as we get older we tend to get more linear and more regular about specific kinds of movements," he says, and wonders whether reduced adult plasticity partly reflects not engaging in behaviors that release these chemicals, rather than the chemicals simply being unavailable [4]. He adds that these principles "also apply to young people" [4].

Where They Agree, and Where They Don't

They agree there is a real early advantage for speech sounds. Chang describes the loss of sensitivity to non-native sounds within a couple of years [1]. Jarvis says speech has a stronger critical period than other skills and that learning without an accent is easier early [2]. Huberman says children learn language through almost passive exposure [3]. Huberman's own struggle with Spanish next to his fluent niece is the everyday version [2].

They agree the adult brain still changes. Jarvis's SRGAP2 point that humans remain "juvenile like" [2], Huberman's account of adult map changes with attention [3], Knudsen's incremental and high-stakes findings [4], and Jarvis's adult stuttering therapy [2] all show adult plasticity in speech-relevant systems.

They agree the rules differ. Children learn from mere exposure; adults need focus, alertness and effort [3]. Jarvis frames the adult disadvantage as "starting from first principles" [2]; Huberman frames it as needing a neurochemical gate opened deliberately [3].

They describe the timing differently. Chang's phoneme window closes within a couple of years [1]. Jarvis links language difficulty to missing exposure before puberty [2]. Huberman marks age 25 for the shift from passive to effortful plasticity, and puberty for the near-end of new neurons [3]. These are not really contradictions: they describe different processes, from early sound tuning, to broad language acquisition, to general plasticity, each with its own clock.

They differ on how "closed is closed. Hubel and Wiesel thought a missed window could not be rescued; Huberman says they were wrong [3]. Jarvis still stresses that we still go through the critical periods like they all do" and that the brain deliberately locks in what it learns [2]. Chang shows the window's timing can be shifted by the sound environment, at a cost [1]. The best reading is that windows narrow rather than slam shut, and that sound-level tuning (accent) is where the narrowing matters most.

They differ on whether bilingual children have an edge later. Jarvis says the literature commonly finds it, but notes some researchers argue against it [2]. His explanation rests on retained phonemes rather than a more plastic brain [2].

Why It Works

Why windows close. Structured sound shapes the young auditory cortex, and maturation narrows it toward the sounds it hears [1]. Jarvis argues closure protects stored knowledge because the brain can only hold so much and survival requires not forgetting [2]. Huberman adds that older brains have less room between neurons and add almost no new neurons after puberty [3].

Why adults can still change. Humans retain more immature speech circuits thanks to an extra SRGAP2 gene copy [2]. In adults, attention plus alertness releases epinephrine and acetylcholine that mark specific circuits, and sleep then strengthens them [3]. Errors trigger that chemistry, and dopamine arriving with partial success accelerates it [4]. Small increments keep errors within what the adult system can absorb, while high stakes push the system toward child-like rates of change [4].

Why accents are hardest. Accents depend on finely tuned perception of phonemes, which narrows early [1][2], and on precise motor control of lips, jaw and larynx in the precentral gyrus [1]. A new language with unfamiliar phonemes forces the adult to start from first principles [2]; multilinguals who kept a wider phoneme set have less to rebuild [2].

Why exposure alone fails adults. Children learn from exposure because their brains are plastic by default [3]. Adults need the attention gate; in Recanzone's experiment, the same touch input changed touch maps only when people attended to it [3]. Chang's point that live human interaction matters more than recordings fits this [1].

Where It Applies, and Where It Doesn't

The early advantage applies most to the sound system: distinguishing and producing phonemes, and speaking without an accent [1][2]. It applies to children raised with two languages, who can merge them in a way adults can't [2], and to learners starting a language entirely from scratch [2].

It applies less to what adults can learn with deliberate practice: new maps and skills when attention, alertness, errors, increments and sleep are in place [3][4]. It doesn't mean adults are locked out of language; Huberman explicitly includes learning a new language in what attention-driven plasticity makes possible at any age [3]. Nor does being multilingual guarantee easier later learning, given the dissent Jarvis mentions [2].

The sources' evidence also has limits. Chang's white noise findings are in rats and don't tell us whether nighttime white noise affects babies [1]. The prism and drum experiments involved sensory and motor maps, not language directly, so applying them to language is Huberman's inference [3][4]. Jarvis's movement-for-cognition advice rests on personal experience [2]. And no source claims a specific age after which an accent is impossible to reduce.

For children with hearing loss, speech delays or a stutter, the sources point to early attention and therapy; a qualified audiologist or speech-language therapist is the right person to guide that.

What To Do With It

1. Expect different rules, not a closed door. As an adult you won't learn from passive exposure the way a child does; plan for focused effort followed by rest [3].

2. Learn in short, focused bouts on small pieces. Avoid cramming; stack small errors the way Knudsen's prisms shifted by 7, then 14, then 28 degrees [4]. Work in roughly 90-minute blocks, expecting focus to arrive after 10 to 15 minutes [4].

3. Stay in the error zone. Aim for 7 to 30 minutes of making mistakes each session, and tell yourself the errors are good for learning [4]. Agitation while focusing is a sign you're doing it right [3].

4. Create a real reason to learn. Tie the language to something that matters, such as a relationship, a job or a move, since high contingency speeds adult plasticity [4].

5. Prioritize live conversation over recordings. Chang says real human interaction matters more, and immersion helps [1]; keep exposure going, because unused language can be lost [1][2].

6. Train the sounds deliberately. Because accents rest on phoneme perception and articulatory control, give special attention to the sounds your native language lacks; you are starting those from first principles [1][2].

7. Sleep and rest after hard sessions. Plasticity happens during sleep; a 20-minute NSDR session or short nap after learning can help [3].

8. Arrive in the right state. Before a session, use a physiological sigh or panoramic vision if anxious, and sleep, NSDR, caffeine or faster breathing if tired [4]; spend 60 to 120 seconds focusing your eyes on your material to engage attention [3].

9. Keep moving and keep speaking. Jarvis recommends regular dancing, walking or running and practicing speech or singing to keep neighboring movement and speech circuits in use [2]; Huberman suggests varied movement and balance challenges may amplify plasticity [4].

10. Give children rich, structured sound early. Talk with them, expose them to more than one language during the early years if you can, and favor natural sounds over constant masking noise [1][2].

Sources

1. Dr. Eddie Chang: The Science of Learning & Speaking Languages · Huberman Lab
2. Dr. Erich Jarvis: The Neuroscience of Speech, Language & Music · Huberman Lab
3. How to Focus to Change Your Brain · Huberman Lab
4. How to Learn Faster by Using Failures, Movement & Balance | Huberman Lab Essentials · Huberman Lab